Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
Nvidia CEO Jensen Huang said he would stop supplying chips to a customer developing unsafe AI software, adding he doesn't think current concern about the technology is warranted. Anthropic CEO Dario Amodei has proposed an industrywide slowdown.
Since July, OpenAI, Anthropic, Meta and Google have each disclosed incidents of models accessing external systems without authorization, including OpenAI's hack into Hugging Face. Experts tell CSO the labs rejected their cybersecurity advice, lumping it under "AI safety."
SubQ Snap reads up to 384K tokens in a single call, including text and images, and returns a decision in under a second at 100K tokens. Pricing is $0.05 per million input tokens, cached or uncached, with free output.
ComfyUI v0.40.0 adds support for Qwen-Image 2.1 Turbo, an 8-step open-source image model, per the official changelog and GitHub release.
Cohere's controlled testbed study finds teams of coding agents can match frontier results at roughly 30% of the cost. It cites prior runs: 16 Claude agents wrote a 100,000-line C compiler for ~$20,000 in API costs, and Bun was ported from Zig to Rust in 11 days with 64 parallel Claude agents.
Drama 3 Preview is the new #1 open-weights text-to-speech model on the Artificial Analysis TTS leaderboard, also leading its Pronunciation Robustness board ahead of Eleven v4 and Eleven v4 Turbo. It supports 83 languages, multi-speaker dialogue, and audio inpainting to fix a single word without a retake.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox